United StatesEducation2025 – ongoing

University of Wisconsin–Madison

Chat/ECL assistant for qualitative research analysis

Wisconsin researchers released a citation-linked AI assistant for just over 5,000 research documents

In progressCompany's own claimOfficial ↗

Problem

An education research team needed shared access to a large qualitative dataset. Different groups collected and analyzed the material.

Approach

Researchers began building Chat/ECL with Gemini in summer 2025 and released it to the full team in spring 2026. After early hallucinations, the developer added planned, repeated searches across raw and coded documents. Answers included confidence information and links to underlying evidence.

Results

just over 5,000documents in the research dataset, per the universityCompany's own claim[1]

Takeaway

The redesign traded speed for more systematic searches and traceable evidence. The team had not reported measured research-time savings.

Sources

From 5,000 Documents to Clear Insights: WCER Researchers Share Lessons From Building Their Own AI Assistant
Wisconsin Center for Education Research17 Jun 2026Company publication

Source summaries

[1]Wisconsin Center for Education Research17 Jun 2026

WCER described Chat/ECL's development and release to researchers. Early errors prompted a slower evidence-gathering process with citations; the account reported dataset size rather than measured productivity gains.

Full details are in the original sources.

Added 1 Nov 2026 · Updated 1 Nov 2026